Anthropic Claude outage of November 13, 2026: claude.ai, Claude Code and Console errors resolved in two hours

Anthropic reported elevated errors across claude.ai, Claude Code, Claude Console and Claude Cowork on November 13, 2026, and marked the incident resolved at 11:58 UTC — exactly two hours after investigation began at 09:58 UTC. The incident, tracked as nn11wk485zfn on status.claude.com, drew 464 points and 373 comments on Hacker News, and the discussion was less about the outage itself than about how much daily work now routes through a single model vendor.

Official incident timeline

Anthropic posted three updates to the status page, moving from Investigating to Monitoring to Resolved in two hours.

Time (UTC) State Status page text
09:58 Investigating We are investigating an issue with users authenticating to Claude.ai and Claude Code.
11:05 Monitoring A fix has been deployed for the issue affecting Claude services and we are monitoring for recovery.
11:58 Resolved The issue affecting claude.ai and Claude Cowork has been resolved.

Two details in the notice are worth flagging. The initial investigation update frames the problem specifically as users authenticating to claude.ai and Claude Code, which points at identity and login rather than at model inference. The final resolved note names only claude.ai and Claude Cowork, even though the incident listing itself named claude.ai, Claude Console (platform.claude.com), Claude Code and Claude Cowork as affected.

Which surfaces were affected

The status page listed four affected services: claude.ai, Claude Console at platform.claude.com, Claude Code, and Claude Cowork. The failure mode users described in the Hacker News thread was consistent with the authentication framing: sessions demanded re-login, and in some cases the re-login itself failed.

Many Claude session suddenly require a re-login, which first worked but now gives OAuth request failed Temporarily unable to authenticate. Please retry.

— user xamde, describing the OAuth re-authentication failure

Other reports from the thread described different symptoms at the edge:

  • Cloudflare interstitials. One commenter saw a 403 Cloudflare attention-required page but was able to finish the session, then switched to a competing tool for a while.
  • Connector and MCP breakage. One commenter noted the incident took down MCP servers, which Anthropic labels Connectors on the status page, and argued this hit business customers using Claude to reach SaaS apps, not only developers writing code.
  • Inconsistent Claude Code behavior. One user in Toulouse, France reported one instance unable to connect and another that appeared to have crashed but kept pretending to work; reconnecting woke the stalled instance.
  • Unaffected API traffic for at least one team. One commenter said their application logs showed zero errors from Anthropic's API that day, suggesting the API path was not uniformly degraded.

The Microsoft and Cloudflare connection is still unconfirmed

The Hacker News thread repeatedly speculated that the Claude incident was collateral from a larger Microsoft outage the same day, but no confirmation from Anthropic appears in the source material.

Is Claude on Azure? GitHub also seems affected, could seem like all of this is related to the huge Microsoft outage

— user Tehnix

Is there indication this is really a separate issue or just Claude not being able to reach Microsoft Entra for whatever periodic requests confirm the user is authenticated (and maybe this page even simply being automatic reflection that those requests are errorring frequently)? Are people who don't use M365 experiencing issues here?

— user JulianWasTaken, proposing the most specific mechanism: Claude failing periodic Entra ID token checks

Other commenters raised the same possibility less specifically, asking whether Anthropic runs infrastructure on Azure, noting a possible link to Microsoft, or pointing at unrelated Cloudflare outages in the same minutes. None of these threads were resolved inside the discussion, and Anthropic's status updates do not state a root cause.

The incident doubled as a dependency audit

The most-quoted comment in the thread was not about Anthropic at all — it was about how much of the workday now stops when a model provider stops.

Previous outages of GitHub, AWS and so on didn't get that much traction. I guess anecdotal but people nowadays are incredibly addicted to LLMs, to the point where they cant or dont want to work when there's an outage

— user slekker

The severity paradox came up too: when an outage is total enough, individual accountability drops because nobody can be blamed for a vendor-wide failure. One commenter observed that the effective severity of such events decreases precisely as the blast radius widens, since everyone simply shrugs.

A longer comment described the uneven adoption inside a single company: roughly 20 percent of developers are doing everything with Claude at very high output, about 10 percent refuse to touch it, and the rest fall somewhere in between — with managers suggesting a return to manual work during outages.

It is incredible to watch this miracle tech hit so fast and have everyone's understanding of it be so different across the board.

— user mr_roboto

Two hours is fast by comparison

The incident ran from 09:58 to 11:58 UTC. One commenter contrasted this with other major outages that lasted far longer, and still judged it embarrassing for a heavily relied-upon service while asking whether Claude itself helped diagnose the problem.

Apparently it was resolved in 2 hours. That seems faster than usual, compare to Outlook (down for a day) and GitHub. It's still embarrassing for a heavily-relied service, but better, and I wonder if Claude (or another LLM) helped diagnose and fix the outage?

— user armchairhacker

The incident notice itself drew criticism

A separate line of complaint targeted the status page rather than the outage. One commenter who was using Claude Code on Friday morning said the notice left them unable to tell what the issue was or how they were affected, and treated that as a reason to accelerate a move away from the vendor.

Claude Code is good. But if when it has problems it issues notices like this, being good is not enough

— user worik

The critique is specific and reproducible from the source: the three posted updates describe states, not causes, and the resolved message narrows the affected-service list without explaining why.

Practical takeaways for teams depending on Claude

Four concrete lessons emerge from the incident and the discussion.

  1. Authentication is a distinct failure domain from inference. The outage presented as re-login loops and OAuth failures, so monitoring that only watches model latency and completion error rates will miss this class of incident.
  2. Connectors and MCP servers are part of the blast radius. Business integrations that reach SaaS systems through Claude were reported down, which extends the impact beyond coding workflows.
  3. Partial degradation is common. Some sessions failed to authenticate while other teams' API logs stayed clean, so a single status-page signal does not describe every customer's experience.
  4. Fallback tooling is a live operational question. Multiple commenters described switching to a competing product during the window, which suggests maintaining a second authenticated path is now a standard continuity measure for LLM-reliant teams.

Open questions the source does not answer

The public record in this incident leaves several things unresolved: what actually broke, whether the degradation originated inside Anthropic or in an upstream identity provider such as Microsoft Entra or a CDN, whether the API itself was affected at any point, and why the resolved notice listed fewer services than the incident. Anthropic's updates state status changes but no root cause.

Sources

Related